This needs to be un-flagged as solved since it is still happening after the post 195 for some, and the hunt for the squareness of the carriage is under scrutiny.
Anyhow - this issue will be an ongoing thing for a while. We thought long and hard on the smooth pulley, then someone finally did a test and disproved that one.
Then it was a calibration thing, kind of helped but not a cure
Now, as mentioned above squareness and belt tension is the target.
Someone will get around to a stepper motor change (this printer does not like TL smoothers at all, which should have worked - caused a feedback loop and the carriage sit and vibrate). Wondering if it is running on a PWM.